© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 17032
    Prev
    Next
    INDEX
    Explanations

    instances where something ends up happening, resulting in a specific outcome

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    333
    +0.09
    0.4%
    635
    +0.08
    0.4%
    727
    +0.08
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    206
    +0.09
    0.02
    727
    +0.08
    0.02
    553
    +0.08
    0.01
    Negative Logits
    grass
    -0.71
     Awareness
    -0.60
    ongyang
    -0.59
    aic
    -0.58
     Warning
    -0.57
     Wire
    -0.57
     indication
    -0.57
     assurance
    -0.57
    minecraft
    -0.57
     Intelligent
    -0.56
    POSITIVE LOGITS
     settling
    0.86
     deleting
    0.85
     getting
    0.84
     owning
    0.84
     needing
    0.83
     liking
    0.81
     being
    0.79
     marrying
    0.78
     having
    0.77
     encountering
    0.77
    Activations Density 0.033%

    No Known Activations